The US Department of Justice has taken a clear stance in the class-action lawsuit filed by the New York Times against OpenAI and Microsoft, siding with AI companies. In a court filing, the DOJ argues that training large language models (LLMs) on copyrighted text does not constitute copyright infringement—because there is a legal distinction between copying for training and what the model actually outputs.
The New York Times sued OpenAI and Microsoft in late 2023, claiming that millions of NYT articles were used without permission to train models like GPT-4. The newspaper demanded billions of dollars in damages and the destruction of models trained on its content. The case is widely regarded as a bellwether for how courts will handle copyright and AI training going forward.
The Essentials
- DOJ Position: Training on copyrighted material is fair use because the output "often if not always" lacks substantial similarity to the originals
- Core Argument: During training, entire works are copied but never made public—fundamentally different from the end product
- Scale Problem Overlooked: The DOJ compares AI training to a teenager copying Hemingway to learn—ignoring commercial mass production
- Internal Contradiction: The US Copyright Office had recently argued in a report that AI training at this scale and speed exceeds fair use limits
The Justice Department's Logic: Learning vs. Publishing
The DOJ uses an intriguing analogy: Joan Didion copied Hemingway stories as a teenager to understand his sentence structure. Under this logic, she would later face liability whenever she published—which would be absurd. The argument: you shouldn't have to pay when you later draw on a book to write something new.
The central thesis: there is a difference between the internal training process (where copies are made but never published) and what the model subsequently produces. A blanket damages theory conflating both is legally flawed.
The Copyright Office Disagrees
The US Copyright Office had argued precisely the opposite in its report. The agency emphasized that AI works with perfect copies and generates content at a speed and scale far beyond human creation. Commercial applications competing with original works in existing markets exceed fair use protections.
Particularly striking: according to reports, the Copyright Office director was fired by the Trump administration shortly after the report's release. The DOJ filing now signals a clear pro-AI course for the new administration.
What This Means for German Companies
This development has implications beyond the US. If American courts confirm fair use for AI training, it will create pressure on European regulation—even though EU AI Act and copyright frameworks are structured differently. German AI companies and users should monitor how this case unfolds. Simultaneously, a key question remains: how will the EU define its own position on AI training with protected material? The issue is not only legal but also strategically competitive.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
